APPLICANT'S ABSTRACT: This project's primary aim is to evaluate four youth alcohol prevention measures: (1) the 21-minimum drinking age, (2) zero tolerance for alcohol in drivers below the legal drinking age, (3) night driving curfews for youth (often implemented through graduated or provisional licensing laws), and (4) alcohol taxation (viewed as a means to reduce alcohol consumption by youth). It also will evaluate wether .08 legal driver blood alcohol limits differentially affect youth. Existing evaluations of these measures are surprisingly weak. The evaluation will use three kinds of time series data: (1) national Fatal Analysis Reporting System (FARS) data about all fatal highway crashes and their alcohol involvement, (2) national Vital Statistics data about all U.S. deaths and their causes, and (3) hospital discharge system (HDS) data from seven states describing hospital utilization by cause. These data sets all include demographic data. The project has four secondary aims: (1) To develop, validate, and apply a model for estimating cause of hospital-admitted injury from patient demographics and diagnoses. (This model is needed because HDS data only have been cause-coded relatively recently); (2) To compare the strengths and weaknesses of the three types of data for evaluating alcohol prevention measures; (3) To estimate if the 21-minimum drinking age is associated with an increase in binge drinking (proxied by acute alcohol poisoning); and (4) To apply existing models that convert the impacts on health services utilization into estimated cost savings. We will evaluate the effects of policy interventions, alternatively by examining and comparting their impacts on injury fatalities and upon hospital-admitted cases. We will pose and answer the questions of wether patterns of fatal injuries match those of hospitalized (=admitted) injuries and what are the implications of the findings for hospital utilization and health care costs. The FARS and Vital Statistics model will span 20 years x 4 Quarters/Yr. X 50 states x one single- year age cohort for each of the 13 single-year age cohorts (ages 14 to 26). We will build "extended families of regression models" for a variety of injury categories and estimate the models with each of the datasets. Our sample includes observations for different states over time. To take advantage of both the cross sectional variation across states and variation over time, we will assess the impacts of the policy interventions described earlier in this proposal by estimating coefficients of pooled cross-sectional (multiple) time series models. This family of models consists of sophisticated regression techniques which differ principally in the way in which error structures and intercepts are specified. By employing generalized least squares, we can easily overcome the shortcomings of the application of OLS methods to pooled cross-sectional time series data. The proposed research will greatly improve our understanding of the impacts of these youth-targeted alcohol measures by separating the fatality impacts of the focus measures from a variety of all-age measures and identify their impacts on health services utilization.